The association between bone mineral density gene variants and osteocalcin at baseline, and in response to exercise: the gene SMART study
This is the Accepted version of the following publication
Hiam, Danielle, Voisin, Sarah, Yan, Xu, Landen, Shanie, Jacques, Macsue, Papadimitriou, Ioannis D, Munson, Fiona, Byrnes, E, Brennan-Speranza, Tara, Levinger, Itamar and Eynon, Nir (2019) The association between bone mineral density gene variants and osteocalcin at baseline, and in response to
exercise: the gene SMART study. Bone, 123. pp. 23-27. ISSN 8756-3282
The publisher’s official version can be found at
https://www.sciencedirect.com/science/article/pii/S8756328219300900
Note that access to this version may require subscription.
Downloaded from VU Research Repository https://vuir.vu.edu.au/38200/
1 The association between bone mineral density gene variants and osteocalcin at baseline, 1
and in response to exercise: the Gene SMART study.
2
Danielle Hiam1, Sarah Voisin1, Xu Yan1, Shanie Landen1, Macsue Jacques1, Ioannis D.
3
Papadimitriou1, Fiona Munson1, Elizabeth Byrnes2, Tara C Brennan-Speranza3, Itamar 4
Levinger*1,4, Nir Eynon*1,5 5
1 Institute for Health and Sport (iHeS), Victoria University, Melbourne, Australia.
6 7
2 PathWest QEII Medical Centre, Perth, Australia 8 9
3 Department of Physiology, University of Sydney, Sydney, NSW, Australia 10 11
4 Science (AIMSS), Department of Medicine-Western Health, Melbourne Medical School, The 12
University of Melbourne, Melbourne, VIC, Australia 13
14
5 Murdoch Childrens Research Institute, Melbourne, Australia 15
16 17 18
* Itamar Levinger and Nir Eynon are sharing senior authorship 19
20
Address for correspondence:
21
Associate Professor Nir Eynon 22
Institute for Health and Sport (iHeS), 23
Victoria University 24
PO Box 14428 25
Melbourne, VIC 8001, Australia.
26
Tel: (61-3) 9919 5615, Fax: (61-3) 9919 5532, E-mail: [email protected] 27 28
29 30
31 32 33 34 35 36 37 38
2 Abstract
39
Introduction: Osteocalcin (OC) is used as a surrogate marker for bone turnover in clinical 40
settings. As bone mineral density (BMD) is largely heritable, we tested the hypothesis that a) 41
bone-associated genetic variants previously identified in Genome-Wide Association Studies 42
(GWAS) and combined into a genetic risk score (GRS) are associated with a) circulating levels 43
of OC and b) the changes in OC following acute exercise.
44
Methods: Total OC (tOC), undercarboxylated OC (ucOC), and carboxylated OC (cOC) were 45
measured in serum of 73 healthy Caucasian males at baseline and after a single bout of high- 46
intensity interval exercise. In addition, genotyping was conducted targeting GWAS variants 47
previously reported to be associated with BMD and then combined into a GRS. Potential 48
associations between the GRS and tOC, ucOC and cOC were tested with linear regressions 49
adjusted for age.
50
Results: At baseline none of the individual SNPs associated with tOC, ucOC and cOC.
51
However, when combined, a higher GRS was associated with higher tOC (β = 0.193ng/mL;
52
p=0.037; 95% CI = 0.012, 0.361) and cOC (β = 0.188ng/mL; p=0.04; 95% CI = 0.004, 0.433).
53
Following exercise, GRS was associated with ucOC levels, (β = 3.864ng/mL; p-value = 0.008;
54
95% CI = 1.063, 6.664) but not with tOC or cOC.
55
Conclusion: Screening for genetic variations may assist in identifying people at risk for 56
abnormal circulating levels of OC at baseline/rest. Genetic variations in BMD predicted the 57
ucOC response to acute exercise indicating that physiological functional response to exercise 58
may be influenced by bone-related gene variants.
59 60 61
3 Key words: Osteocalcin, SNP, exercise, bone turnover, biomarkers, bone mineral density.
62 63
Introduction 64
The primary function of the skeletal system is to provide mechanical support of the body, 65
respond to outside mechanical forces, and is a reservoir for normal mineral metabolism [1, 2].
66
Throughout the lifespan the skeleton undergoes continuous bone remodelling, it is tightly 67
controlled by osteoclasts and osteoblasts which balance bone removal and bone formation [3].
68
In-vivo bone remodelling is difficult to assess, therefore circulating bone turnover markers 69
(BTMs) are commonly used in a clinical setting as a surrogate measure for bone metabolism 70
and give an indication of bone turnover [4, 5].
71
Osteocalcin (OC), a BTM, is secreted by osteoblasts into the extracellular matrix and is 72
important in bone matrix formation, mineralization and maintenance [6, 7]. OC can be post- 73
translationally modified by gamma (ƴ)-carboxylation at one or more of the three (17, 21 and 74
24) glutamic acid residues and found in two different forms, carboxylated (cOC) and 75
undercarboxylated OC (ucOC). While they share a similar tertiary structure they are thought 76
to have different biological functions. The undercarboxylated form (ucOC) was implicated in 77
energy metabolism and cardiovascular health [4, 8-10]. UcOC may also play a role in bone 78
health as higher levels of ucOC have been found to be associated with a higher risk of hip 79
fracture [11-13]. While cOC, where all three glutamic acid residues are carboxylated, is 80
considered predominantly a protein of the bone matrix [6, 9]. After carboxylation 81
conformational changes occur in OC, increasing the affinity for calcium ions [9]. Calcium helps 82
to stabilize OC and facilitates the binding of OC to the surface of hydroxyapatite, the bone 83
mineral component of the bone and is closely aligned with the bone mineral density (BMD) [4, 84
6].
85
4 BMD is a parameter used in the identification of people at risk for osteopenia and osteoporosis 86
[14, 15]. BMD has a high genetic heritability with contributions from environmental factors 87
such as exercise and nutrition [16-18]. A recent Genome-Wide Association (GWAS) study 88
identified nine single nucleotide polymorphisms (SNPs) that are associated with BMD (p< 5 89
X 10-6) [16]. Environmental factors such as acute exercise can also affect BTMs by 90
mechanically loading the skeleton [19-21]. We have also previously shown that the ACTN3 91
R577X common SNP is associated with serum levels of tOC in men with serum levels higher 92
in the ACTN3 XX genotype (α-actinin-3 deficiency) compared to RR and RX at baseline [22].
93
This illustrates both genetic and environmental factors can influence bone turnover and the 94
associated markers [23, 24]. Any imbalances to this process can cause bone loss and a reduction 95
in bone strength leading to common bone disorders such as osteopenia and osteoporosis [25].
96
However, it is currently unclear whether bone related genetic variants are associated with levels 97
of bone turnover, nor if genetic variants can alter the response of BTMs following acute 98
exercise. This could be clinically important in identifying people at a younger age at risk of 99
bone related disorders. Therefore, we tested the hypothesis that the GWAS SNPs that were 100
previously identified to be associated with bone-related phenotypes [16], can predict 101
circulating tOC, ucOC and cOC at baseline, and following an acute bout of High-Intensity 102
Interval Exercise (HIIE).
103
Materials and methods 104
Participants 105
This study is a part of the Genes and Skeletal Muscle Adaptive Response to Training (Gene 106
SMART) study. The detailed methodology has previously been published [26]. Briefly, 107
seventy-three apparently healthy, Caucasian men (age = 31.4 years ± 8.2; BMI = 25.2 kg/m2 ± 108
3.2) participated in the study following a written informed consent. Participants attended the 109
5 VU laboratory on 2 separate occasions, to perform two graded exercise tests, and on one 110
occasion for the blood sampling and the acute exercise intervention. Volunteers were excluded 111
if they had a bone disease, were taking hypoglycaemic medications, warfarin or vitamin K 112
supplementation, or medications that affect bone metabolism, insulin secretion, or sensitivity.
113
Further, participants with known musculoskeletal or other conditions that prevent daily activity 114
were excluded from the study. This study was approved by the Human Ethics Research 115
Committee at Victoria University (HRE13-223) and all participants provided written informed 116
consent.
117
Aerobic Capacity (Graded exercise test) 118
Aerobic capacity was assessed by a graded exercise test (GXT) to establish peak power output 119
(Wpeak) and lactate threshold (LT). Briefly the test consisted of 4 minute exercise stages, 120
separated by 30 second rest periods until voluntary exhaustion. Capillary blood samples were 121
collected and analysed by the YSI 2300 STAT Plus system (Ohio, USA) at the end of each 4 122
minute stage and immediately after exhaustion to establish lactate (LT) concentration. LT was 123
calculated by the modified DMAX method as previously described [26].
124
Nutrition consultation 125
Each participant was provided with an individualised pre-packaged diet 48 hours prior to 126
providing the blood samples to standardise diet across the participants and minimise the effects 127
of this confounding factor [27, 28]. The content of the diets were based on the current 128
Australian National Health and Medical Research Council (NHMRC) guidelines. Participants 129
were asked to abstain from food, caffeine and alcohol 12 hours prior to blood collection.
130
Acute exercise session 131
6 Participants completed a single session of High Intensity Interval Exercise (HIIE) on a cycle 132
ergometer (Velotron®, Racer Mate Inc.). The acute exercise session consisted of 8 X 2min 133
intervals at 40% of [(Wpeak-LT) + LT] with 1 minute active recovery intervals at a power of 134
60W.
135
Serum osteocalcin measurements 136
Before acute exercise (baseline), immediately after the acute exercise, and 3 hours post 137
exercise, venous blood samples were collected via venepuncture or cannulation in BD SST 138
Vacutainers (Becton and Dickson Company, USA). All participants abstained from food, 139
caffeine and alcohol 12 hours prior to blood collection in the morning to control for variation 140
in serum levels of OC. They were left at room temperature (10 mins) before being centrifuged 141
at 3500 rpm for 10mins at 4°C. Serum was collected and stored at -80°C. tOC was measured 142
using an automated immunoassay (Elecys 170; Roche Diagnostics). This assay has a sensitivity 143
of 0.5 µg.L-1 with an intra-assay precision of 1.3%. We measured ucOC by the same immuno- 144
assay after absorption of carboxylated OC on 5mg/ml hydroxyl-apatite slurry as described by 145
Gundberg et al. [29]. cOC was calculated by the subtracting the ucOC from the tOC.
146
Circulating tOC, ucOC and cOC were measured at baseline (before acute exercise), 147
immediately after, and 3 hours post exercise. The peak OC, ucOC and cOC was considered the 148
maximal concentration immediately after or 3 hours post exercise 149
Genotyping 150
Genomic DNA was extracted from residual blood samples from BD Vacutainer EDTA tubes 151
using the MagSep Blood gDNA kit (0030 451.00, Eppendorf, Hamburg, Germany) [26]. The 152
genotype of each SNP was assessed by the Australian Genome Research Facility (AGRF). We 153
chose the SNPs based on previous literature from a GWAS study assessing BMD [16]. The 154
SNPs, locus, type, effect allele and closest gene are described in supplementary table 1. Based 155
7 on the Moayyeri et al GWAS, nine genetic variants known to be associated with broadband 156
ultrasound attenuation (BUA), that estimates BMD, were used to calculate genetic risk scores 157
(GRS).The SNPs were measured by MassARRAY® combined with iPLEX® chemistry 158
(Agena Bioscience).
159
Statistical analysis 160
We conducted linear regressions to test for potential associations between individual SNPs and 161
tOC or cOC levels, using an additive genetic model (i.e. homozygotes for non-effect allele 162
coded as 0; heterozygotes coded as 1; homozygotes for effect allele coded as 2). The GRS was 163
calculated by coding each SNP with the number of effect alleles (0, 1 or 2), multiplying this 164
with the published regression coefficient (beta) [16] then summed up to obtain a weighted GRS.
165
Linear regressions were conducted to test whether the GRS was associated with tOC, ucOC or 166
cOC. The distribution of tOC, ucOC and cOC were checked for normality visually using 167
histograms, and tOC, ucOC and cOC were log-transformed to meet normality assumptions. In 168
the linear regressions, we used log-transformed tOC, ucOC or cOC as the dependent variable;
169
and GRS and age as the independent variables. We also tested BMI, and fitness parameters 170
(VO2peak, lactate threshold and peak power) as additional covariates, but as they did not 171
significantly associate with levels of tOC, ucOC or cOC they were removed from the model.
172
To investigate the response of tOC, ucOC and cOC to exercise and potential associations with 173
the GRS, we used the delta change of baseline to peak before and after the acute exercise 174
session. Where required p-values from the statistical analyses were adjusted for multiple testing 175
using the false discovery rate (FDR) [30], and q-values<0.05 were deemed significant. Post- 176
hoc power analysis was conducted in R using the pwr package using effect sizes and notations 177
from Cohen J [31].
178 179
8 Results
180
Participants’ characteristics and the tOC, ucOC and cOC responses to exercise are described 181
in Table 1. A small, but significant (4.6%, p<0.05) increase was observed for tOC, and ucOC 182
(10.1%, p<0.01) but not cOC, following exercise (Table 1).
183
Table 1- Participant characteristics (n=73) 184
Baseline Age (years) 31.4 ± 8.2 BMI (kg.m-2) 25.2 ± 3.2 VO2peak (mL.kg-1.min-1) 47.2 ± 8.0 Lactate threshold (W) 206.7 ± 55.3
Wpeak (W) 294.9 ± 64.7
Circulating Osteocalcin Peak p-value
tOC (ng/ml) 30.5 ± 10.9 31.9 ± 10.65 p=0.004 cOC (ng/ml) 18.7 ± 8.2 19.1 ± 7.8 p=0.372 ucOC (ng/ml) 11.9 ± 4.3 13.1 ± 4.6 P<0.01 BMI, Body Mass Index; VO2peak, Peak oxygen uptake during graded exercise test; Peak, peak 185
level of tOC, ucOC and cOC at either immediately post or 3 hours post exercise; tOC, total 186
osteocalcin; cOC, Carboxylated OC; ucOC, under-carboxylated osteocalcin; W, watts. Values 187
are mean ± SD.
188
Individual SNPs were not associated with tOC, ucOC and cOC 189
We first assessed the contribution of each individual SNP to the variance in tOC, ucOC and 190
cOC, but no significant associations were found (Table 2).
191
9 Table 2- Individual SNP regression analysis with tOC, ucOC and cOC.
192
tOC cOC ucOC
SNP ID Β
(ng/ml) p- value
FDR q- value
Post- Hoc Power
(%)
Β (ng/ml)
p-
value FDR
Post- Hoc Power
(%)
Β (ng/ml)
p- value
FDR q- value
Post- Hoc Power
(%) rs7741021 0.08 0.82 0.92 5.3 0.144 0.71 0.71 5.6 0.124 0.889 0.889 0.05
rs4869739 -0.19 0.54 0.69 7.9 -0.218 0.57 0.64 7.5 -0.385 0.655 0.75 0.066
rs3020331 0.29 0.38 0.68 11.6 0.348 0.39 0.64 11.6 0.449 0.616 0.75 0.072
rs2982552 0.80 0.02 0.19 55 0.82 0.059 0.3 39.3 1.75 0.07 0.63 0.075
rs2908007 0.22 0.13 0.39 27 0.232 0.18 0.54 21.3 0.309 0.423 0.75 0.054 rs597319 0.01 0.98 0.98 0.5 0.217 0.57 0.64 7.6 -0.873 0.299 0.672 0 rs10416265 -0.25 0.49 0.69 8.8 -0.32 0.47 0.64 9.1 -0.429 0.667 0.75 0.14
rs6974574 0.62 0.21 0.47 19.1 0.544 0.37 0.64 11.6 1.588 0.243 0.67 0.17 rs38664 0.90 0.06 0.28 37.8 1.091 0.068 0.31 37 1.672 0.212 0.67 0.9 The p-values were adjusted for multiple testing using the false discovery rate (FDR). Additive genetic models were used. Effect size correspond 193
to the regression coefficient in the linear models, and is interpreted as the change in tOC, ucOC or cOC (log-transformed) per effect allele at the 194
SNP.
195
10 The GRS was positively associated with tOC, ucOC and cOC
196
As the contribution of each individual SNP may be too small to be detected in only n = 73 197
individuals, we calculated a GRS to increase statistical power (see Material & Methods). Age 198
was negatively associated with tOC (β = -0.608ng/ul; p<0.001; 95% CI = -0.017, -0.009), ucOC 199
(β = -0.018ng/ml; p<0.001; 95% CI= -0.027, -0.009) and cOC (β = -0.607ng/ml; p<0.001; 95%
200
CI = -0.017, -0.009). After adjusting for age, higher GRS was associated with higher levels of 201
tOC (β = 0.193ng/ml; p-value = 0.037; 95% CI = 0.012, 0.361) and with higher levels of cOC 202
(β = 0.188ng/ml p-value = 0.046; 95% CI = 0.004, 0.433). The GRS explained 6.1% of the 203
variance in tOC and 5.6% of the variance in cOC (Figure 1). ucOC was not associated with 204
GRS (β=0.261ng/ml; p-value = 0.289; 95% CI= -0.226, 0.747) after adjusting for age (Figure 205
206 1).
207 208 209 210 211 212 213 214 215 216
11 217
Figure 1- Regression analysis of genetic risk score with tOC,ucOC and cOC adjusted for 218
age. BMD GRS = Bone Mineral Density Genetic Risk Score. Significance p<0.05 after 219
adjusting for age.
220
12 221
The GRS was associated with exercise-induced peak in ucOC but not tOC or cOC 222
A positive association was identified between ucOC response to exercise and the GRS (β = 223
3.864ng/ml; p-value = 0.008; 95% CI = 1.063, 6.664). The GRS explained 9.8% of the variance 224
in ucOC response to exercise (Figure 2). We were unable to identify an association between 225
the changes in tOC or cOC and the GRS (p=0.617 and p=0.17, respectively).
226
227
Figure 2- Regression analysis of genetic risk score with change in ucOC (Δ) after acute 228
exercise. BMD GRS = Bone Mineral Density Genetic Risk Score. Significance p<0.05 after 229
adjusting for age.
230 231
Discussion 232
We report that a higher GRS is associated with higher tOC and cOC at baseline in healthy 233
young men but not with ucOC. GRS did not appear to predict the change in tOC or cOC 234
13 following exercise. However, a higher genetic risk for lower BMD was associated with the 235
increased concentration of ucOC following exercise, providing a novel insight that BMD 236
genetic variants may play a role in this response.
237
BMD is a complex trait that, in part, relates to heritability and is used as a predictor for future 238
risk to develop osteoporosis. Yet, each individual SNP may only contribute a small amount to 239
the hereditary component of BMD, making this type of analysis not useful in predictive studies 240
[32]. This is shown in table 2 where each individual SNP did not predict change in tOC, ucOC 241
or cOC. GRS combines SNPs identified by GWAS into a score that can evaluate and provide 242
further insights into the genetic contribution to BMD by increasing statistical power to detect 243
these small contributions [32-34]. We used nine SNPs that were associated with BMD from an 244
unbiased GWAS approach in large consortium [16] and found that these SNPs are also 245
associated with the BTM, OC. The genetic variants used in the GRS determined 6.1% and 5.6%
246
of the variability of circulating levels of tOC and cOC, respectively, at rest and 9.8% of the 247
variability of circulating levels of ucOC in response to acute exercise.
248
We report that healthy men who have a higher genetic risk for lower BMD displayed increased 249
circulating levels of tOC. Previous studies have shown that higher levels of serum BTMs, 250
including tOC are associated with higher bone turnover [22, 35, 36]. While OC is 251
conventionally used as a marker of bone formation, literature suggests that it may be a better 252
indicator of overall bone turnover [37, 38]. Studies have shown that osteoporosis, vertebral 253
fractures, and bone loss are associated with increased levels of tOC [39, 40]. Therefore, we 254
provide evidence indicating that genetics may be playing a role in influencing bone turnover, 255
previously shown to be associated with ongoing bone loss or higher risk of fracture [37, 39, 256
40].
257
14 Exercise mechanically loads the skeleton, improves insulin sensitivity and may partly mediate 258
the interaction between muscle, bone and glucose metabolism [9, 20]. Therefore, we explored 259
the hypothesis that BMD gene variants may play a role in tOC, ucOC or cOC response to acute 260
exercise. We found that the genetic risk score was not associated with tOC or cOC in response 261
to acute exercise, suggesting that the genetic variables examined cannot determine the response 262
of tOC or cOC to acute exercise, at least in healthy-young men [22]. We could speculate that 263
as bone formation is a slow process [41], to observe any influence of BMD gene variants on 264
circulating levels of tOC or cOC would require a longer exercise intervention. Interestingly, an 265
increased genetics risk for a lower BMD was associated with increased levels of ucOC in 266
response to exercise. It is not clear why those with increased genetic risks exhibited increased 267
levels of ucOC following exercise, however, previous studies suggest that increased ucOC 268
levels are associated with increased fracture risk [11-13]. In addition, the increase in ucOC 269
following exercise may be due to the increase metabolic demands by skeletal muscle. Indeed, 270
we have previously shown that ucOC has a direct effect on muscle glucose uptake in insulin 271
signalling proteins [9, 42]. We confirmed that ucOC is upregulated after exercise in humans 272
and provide a novel insight that BMD genetic variants may play a role in this response.
273
The current study has some potential limitations. First, it includes a relatively small sample 274
size in the context of genetic studies. Yet, even with a small sample size we were able to detect 275
a significant association of GRS with tOC and cOC. We are confident in our findings, as they 276
support previous findings of a GWAS that identified these SNPs to be associated with BMD 277
[16]. We were underpowered for the individuals SNP analysis (on average tOC- 19.2%, ucOC- 278
0.2% and cOC- 16.7%). However, our power improved greatly with the use of the genetic risk 279
score (tOC- 48%, ucOC- 14.9% and cOC- 44%) illustrating the strength of calculating a GRS 280
for data analysis. We acknowledge that while greater sample sizes are required, our data 281
provides hypothesis generating pilot data that offer interesting novel concepts for future 282
15 analysis. Secondly, we did not assess BMD, therefore a direct assessment of the SNPs with 283
BMD cannot be performed. Finally, we only tested young-healthy males. Future studies should 284
focus on young healthy females as well as older adults who have an increased risk for 285
osteopenia and osteoporosis.
286
In conclusion, screening for genetic variations may assist in identifying people at risk for 287
abnormal circulating levels of OC. Genetic variations in BMD predicted the response of ucOC 288
to acute exercise, but not tOC or cOC, indicating that physiological functional response to 289
exercise may be influenced by bone-related gene variants.
290 291
Acknowledgements 292
A/Prof Levinger was supported by a Future Leader Fellowship (ID: 100040) from the National 293
Heart Foundation of Australia. This research was supported the National Health & Medical 294
Research Council (NHMRC CDF # APP1140644) to Nir Eynon.
295 296
References 297
1. Levinger, I., et al., Multifaceted interaction of bone, muscle, lifestyle interventions and 298 metabolic and cardiovascular disease: role of osteocalcin. Osteoporos Int, 2017. 28(8): p.
299 2265-2273.
300 2. Seeman, E., Bone quality: the material and structural basis of bone strength. J Bone Miner 301 Metab, 2008. 26(1): p. 1-8.
302 3. Gundberg, C.M., J.B. Lian, and S.L. Booth, Vitamin K-dependent carboxylation of osteocalcin:
303 friend or foe? Adv Nutr, 2012. 3(2): p. 149-57.
304 4. Neve, A., A. Corrado, and F.P. Cantatore, Osteocalcin: skeletal and extra-skeletal effects. J 305 Cell Physiol, 2013. 228(6): p. 1149-53.
306 5. Parfitt, A.M., What is the normal rate of bone remodeling? Bone, 2004. 35(1): p. 1-3.
307 6. Hauschka, P.V., J.B. Lian, and P.M. Gallop, Direct identification of the calcium-binding amino 308 acid, gamma-carboxyglutamate, in mineralized tissue. Proc Natl Acad Sci U S A, 1975. 72(10):
309 p. 3925-9.
310 7. Harada, S. and G.A. Rodan, Control of osteoblast function and regulation of bone mass.
311 Nature, 2003. 423(6937): p. 349-55.
312
16 8. Levinger, I., et al., Undercarboxylated osteocalcin, muscle strength and indices of bone health 313 in older women. Bone, 2014. 64: p. 8-12.
314 9. Lin, X., et al., Undercarboxylated Osteocalcin: Experimental and Human Evidence for a Role in 315 Glucose Homeostasis and Muscle Regulation of Insulin Sensitivity. Nutrients, 2018. 10(7).
316 10. Lee, N.K., et al., Endocrine Regulation of Energy Metabolism by the Skeleton. Cell, 2007.
317 130(3): p. 456-469.
318 11. Luukinen, H., et al., Strong prediction of fractures among older adults by the ratio of 319 carboxylated to total serum osteocalcin. J Bone Miner Res, 2000. 15(12): p. 2473-8.
320 12. Szulc, P., et al., Serum undercarboxylated osteocalcin correlates with hip bone mineral 321 density in elderly women. J Bone Miner Res, 1994. 9(10): p. 1591-5.
322 13. Szulc, P., et al., Serum undercarboxylated osteocalcin is a marker of the risk of hip fracture in 323 elderly women. J Clin Invest, 1993. 91(4): p. 1769-74.
324 14. Hernlund, E., et al., Osteoporosis in the European Union: medical management, 325 epidemiology and economic burden. A report prepared in collaboration with the
326 International Osteoporosis Foundation (IOF) and the European Federation of Pharmaceutical 327 Industry Associations (EFPIA). Arch Osteoporos, 2013. 8: p. 136.
328 15. Kanis, J.A., et al., European guidance for the diagnosis and management of osteoporosis in 329 postmenopausal women. Osteoporos Int, 2013. 24(1): p. 23-57.
330 16. Moayyeri, A., et al., Genetic determinants of heel bone properties: genome-wide association 331 meta-analysis and replication in the GEFOS/GENOMOS consortium. Hum Mol Genet, 2014.
332 23(11): p. 3054-68.
333 17. Gaffney-Stomberg, E., et al., Association Between Single Gene Polymorphisms and Bone 334 Biomarkers and Response to Calcium and Vitamin D Supplementation in Young Adults 335 Undergoing Military Training. J Bone Miner Res, 2017. 32(3): p. 498-507.
336 18. Pocock, N.A., et al., Genetic determinants of bone mass in adults. A twin study. J Clin Invest, 337 1987. 80(3): p. 706-10.
338 19. Levinger, I., et al., The effect of acute exercise on undercarboxylated osteocalcin and insulin 339 sensitivity in obese men. J Bone Miner Res, 2014. 29(12): p. 2571-6.
340 20. Levinger, I., et al., The effect of acute exercise on undercarboxylated osteocalcin in obese 341 men. Osteoporos Int, 2011. 22(5): p. 1621-6.
342 21. Mezil, Y.A., et al., Response of Bone Turnover Markers and Cytokines to High-Intensity Low- 343 Impact Exercise. Med Sci Sports Exerc, 2015. 47(7): p. 1495-502.
344 22. Levinger, I., et al., The influence of alpha-actinin-3 deficiency on bone remodelling markers in 345 young men. Bone, 2017. 98: p. 26-30.
346 23. Bjornerem, A., et al., Remodeling markers are associated with larger intracortical surface 347 area but smaller trabecular surface area: a twin study. Bone, 2011. 49(6): p. 1125-30.
348 24. Zhai, G., et al., Genetic and environmental determinants on bone loss in postmenopausal 349 Caucasian women: a 14-year longitudinal twin study. Osteoporos Int, 2009. 20(6): p. 949-53.
350 25. Feng, X. and J.M. McDonald, Disorders of bone remodeling. Annu Rev Pathol, 2011. 6: p. 121-
351 45.
352 26. Yan, X., et al., The gene SMART study: method, study design, and preliminary findings. BMC 353 Genomics, 2017. 18(Suppl 8): p. 821.
354 27. Willems, H.M.E., et al., Diet and Exercise: a Match Made in Bone. Current osteoporosis 355 reports, 2017. 15(6): p. 555-563.
356 28. Starup-Linde, J., et al., Differences in biochemical bone markers by diabetes type and the 357 impact of glucose. Bone, 2016. 83: p. 149-155.
358 29. Gundberg, C.M., et al., Vitamin K Status and Bone Health: An Analysis of Methods for 359 Determination of Undercarboxylated Osteocalcin1. The Journal of Clinical Endocrinology &
360 Metabolism, 1998. 83(9): p. 3258-3266.
361
17 30. Benjamini, Y. and Y. Hochberg, Controlling the False Discovery Rate: A Practical and Powerful 362 Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B
363 (Methodological), 1995. 57(1): p. 289-300.
364 31. Cohen, J., Statistical Power Analysis for the Behavioral Sciences, J. Cohen, Editor. 1988, 365 Academic Press. p. 1-17.
366 32. Ho-Le, T.P., et al., Prediction of Bone Mineral Density and Fragility Fracture by Genetic 367 Profiling. J Bone Miner Res, 2017. 32(2): p. 285-293.
368 33. Kim, S.K., Identification of 613 new loci associated with heel bone mineral density and a 369 polygenic risk score for bone mineral density, osteoporosis and fracture. PLOS ONE, 2018.
370 13(7): p. e0200785.
371 34. Inouye, M., et al., Genomic Risk Prediction of Coronary Artery Disease in 480,000 Adults:
372 Implications for Primary Prevention. J Am Coll Cardiol, 2018. 72(16): p. 1883-1893.
373 35. Dai, Z., et al., Bone turnover biomarkers and risk of osteoporotic hip fracture in an Asian 374 population. Bone, 2016. 83: p. 171-177.
375 36. Bauer, D.C., et al., Biochemical markers of bone turnover, hip bone loss, and fracture in older 376 men: the MrOS study. J Bone Miner Res, 2009. 24(12): p. 2032-8.
377 37. Ivaska, K.K., et al., Release of intact and fragmented osteocalcin molecules from bone matrix 378 during bone resorption in vitro. J Biol Chem, 2004. 279(18): p. 18361-9.
379 38. Garnero, P., et al., Increased bone turnover in late postmenopausal women is a major 380 determinant of osteoporosis. J Bone Miner Res, 1996. 11(3): p. 337-49.
381 39. Delmas, P.D., et al., Assessment of bone turnover in postmenopausal osteoporosis by 382 measurement of serum bone Gla-protein. J Lab Clin Med, 1983. 102(4): p. 470-6.
383 40. Szulc, P., A. Montella, and P.D. Delmas, High bone turnover is associated with accelerated 384 bone loss but not with increased fracture risk in men aged 50 and over: the prospective 385 MINOS study. Ann Rheum Dis, 2008. 67(9): p. 1249-55.
386 41. Katsimbri, P., The biology of normal bone remodelling. European Journal of Cancer Care, 387 2017. 26(6): p. e12740.
388 42. Levinger, I., et al., The effects of muscle contraction and recombinant osteocalcin on insulin 389 sensitivity ex vivo. Osteoporos Int, 2016. 27(2): p. 653-63.
390 391